Practical data acquisition strategy for time-lapse experiments using crosshole GPR and full-waveform inversion

Crosshole ground penetrating radar (GPR) methods are increasingly used in time-lapse studies of flow in the uppermost near subsurface with important implications for our understanding of e.g., water infiltration in the unsaturated zone, and fluid flow in the saturated zone. A particular challenge in such time-lapse crosshole studies is the trade-off between collecting sufficient data to be able to resolve how a tracer moves, and, minimizing the data acquisition time such that the data approximates a static state. We test how dense recording geometries are needed for resolving a gas bubble injected in a highly heterogeneous chalk reservoir analogue using a full-waveform inversion (FWI) approach for modelling the crosshole GPR data. We show that even relatively sparse geometries provide sufficient resolution of the permittivity contrast caused by the gas bubble, provided that the detailed background permittivity structure is known from prior (before gas injection) FWI analysis of densely recorded high-resolution data. The conductivity contrast caused by the gas is more challenging to recover and the resolution suffers to a higher degree when reducing the survey geometry or at higher noise levels. As long as the permittivity change during the time-lapse experiment is the main target, a significant reduction in acquisition time is therefore possible as compared to the time needed to record the background permittivity structure. This reduced acquisition time has important practical implications for time-lapse experiments under realistic conditions. Our results are based on synthetic analysis based on a realistic subsurface scenario closely linked to characterization of heterogeneous chalk reservoirs. However, our findings also have important implications for planning of future time-lapse studies in other settings.